MultiViz: Towards Visualizing and Understanding Multimodal Models

Abstract

The promise of multimodal models for real-world applications has inspired research in visualizing and understanding their internal mechanics with the end goal of empowering stakeholders to visualize model behavior, perform model debugging, and promote trust in machine learning models. However, modern multimodal models are typically black-box neural networks, which makes it challenging to understand their internal mechanics. How can we visualize the internal modeling of multimodal interactions in these models? Our paper aims to fill this gap by proposing MULTIVIZ, a method for analyzing the behavior of multimodal models by scaffolding the problem of interpretability into 4 stages: (1) unimodal importance: how each modality contributes towards downstream modeling and prediction, (2) cross-modal interactions: how different modalities relate with each other, (3) multimodal representations: how unimodal and cross-modal interactions are represented in decision-level features, and (4) multimodal prediction: how decision-level features are composed to make a prediction. MULTIVIZ is designed to operate on diverse modalities, models, tasks, and research areas. Through experiments on 8 trained models across 6 real-world tasks, we show that the complementary stages in MULTIVIZ together enable users to (1) simulate model predictions, (2) assign interpretable concepts to features, (3) perform error analysis on model misclassifications, and (4) use insights from error analysis to debug models. MULTIVIZ is publicly available, will be regularly updated with new interpretation tools and metrics, and welcomes inputs from the community.Preprint. Under review.

Adventures of Trustworthy Vision-Language Models: A Survey

“…It uses four modality attention heads: languageto-vision attention, vision-to-language attention, languageto-language attention, and vision-to-vision attention, allowing it to look at interactions within and between modalities. MULTIVIZ (Liang et al 2022) is another method to analyze multimodal models interpreting unimodal interactions, cross-modal interactions, multi-modal representations, and multimodal prediction. gScoreCAM (Chen et al 2022) studied the CLIP (Radford et al 2021) model specifically to understand large multimodal models.…”

Toward Explainable Affective Computing: A Review

“…While the experiments performed only involve the visual and text modalities due to the high computational cost of the method, user evaluations show that DIME can help researchers determine which unimodal or multimodal contributions are the dominant factors behind the model's prediction. An improvement of DIME aimed at improving its scalability is introduced in MULTIVIZ [183], a tool for analyzing the behavior of multimodal models that scaffold the problem of interpretability into unimodal importance, cross-modal interactions, multimodal representations, and multimodal predictions.…”